Remove EUROMOD comparator facts from Belgium targets#80
Conversation
003979e to
840f1da
Compare
|
Closing after lead review — this PR overshoots the rule it enforces, and the rule is already enforced structurally. The primary-source-only rule governs what may be a calibration target: model outputs must never enter the target surface. That is already guaranteed two ways: (1) every row in this package is typed What this PR deletes is the validation comparator itself, which issue #69's acceptance criteria explicitly require: it feeds the populace#264 EUROMOD-oracle comparison (the independent-engine agreement result). The rows record what JRC published in the 2025 Belgium country report — including the report's own external/administrative columns and ratios — with per-row source URLs and a sha256-pinned artifact. "A publisher asserted X" is a fact under this repo's charter; asserting it as a comparator is exactly the typed role the schema provides. Merging this would also break two currently-green tests ( If a Belgian target profile is ever found binding a euromod-series fact, that binding — not the package — is the bug to fix. |
Summary
Facts
Verification
uv run pytest -q tests/test_belgium_targets.py tests/test_ledger_source_package.pyuv run pytest -q tests/test_ledger_bundle.py::test_build_bundle_writes_merged_consumer_contractuv run ruff checkgit diff --check origin/main...HEADRefs #69